With this great work done in Mayotte by OSM France and especially @SeverinGeo I was wondering how to best tag damage on OpenStreetMap. There are many contradicting informations on the OSM wiki and it would be great to somehow agree a bit better how we do this on OpenStreetMap.
It seems like lots of damage mapping currently on OpenStreetMap is using a lifecycle prefix, especially the destroyed prefix. But in my eyes this does not allow for detailed enough damage mapping, a building is not necessarily complete destroyed and uninhabited.
There is as well this wiki page describing damage mapping. What explains what was used in previous damage mapping projects - which is different than the usage of the lifecycle prefixes.
What was used in Mayotte and some other previous damage mapping projects was the BAR methodology. Which seem to be developed for wind disaster damage assessments. So can this categorization as well be used in different disasters such as an earthquake?
For Cyclone Chido (which is a wind disaster) the BAR methodology was used and the following tags were used:
building=yes
damage:Chido=none OR minimal OR significant OR complete → according BAR methodology
damage:Chido:assessement=2023-03 → where the value is the date of the imagery used or damage assessment done on the ground
damage:Chido:event=TC-2024-000225-COM → where the value is the GLIDE number of the disaster, to be found on this reference site.
source:damage:Chido=CNES / Airbus → where the value is that of the organization that created the imagery or did the survey
And as such I think this makes perfect sense in this situation, but I was wondering how than to adapt it for more conflict situations such as Khartoum in Sudan at the moment. Should we than best leave out the tags of such an event because it is not related to one short event but rather to an ongoing crises?
There are indeed already a lot of initiatives for damage mapping, but there is to my knowledge no real coordinated effort or global database existing at the moment. Also not all the damage databases that are around are of great quality and could use human eyes (on the ground) to verify if it is of good quality. So I think it is especially on these two points where OpenStreetMap brings value; having all data in one global database that is being used already by a lot of first responders and humanitarian organisations + the aspect of human validation and knowledge of local contributors.
Why a more detailed classification? Because it is of importance to understand damage more in detail for example for first responders. Has a neighbourhood or village only buildings that are minimally damaged - than they will not need to go there, if there is a neighbourhood with most of the buildings severely damaged, it is important to go there! As well for other activities more detailed information is useful: based on population numbers we can approximately understand how many people are living in severely damaged buildings and plan for distributions for food and non-food items, being aware of what roads are damaged can ambulances still drive over it or not, as well as for advocacy purpose it is important to understand the impact of damage such as in this example.
However I think I’m not necessarily saying that always we should do full damage assessments on OpenStreetMap, my question rather came up because people are currently mapping damage on OpenStreetMap - which is at the moment rather difficult to use because of all the different tags that are used.
I think in the example of Chido - the tag source:damage:Chido=CNES / Airbus , refers to a satellite image. This was not an import of data from another database into OSM. This tag could be for example as well source:damage:Chido=Bing Imagery or source:damage:Chido=survey
Certainly, but I’m not understanding how it translates to tagging damage as minimal or none. To find areas that were most damaged, look for severe damage tags and compare to overall number of buildings. Or is this intended as a tag to indicate that the building/area was assessed and no damage was seen (as opposed to not knowing if a building was not yet assessed)? If so, I would suggest doing away with distinction between minimal or none.
Also complete damage should probably result in building being tagged demolished:building=* rather than building=*.
A follow-up question is whether these event-specific damage tags are intended to be updated as the damage is repaired, when will they be updated, and who will do it? Having a solid plan for this will help avoid a perception that OSM is being used as a dumping ground for data out of convenience.
I think it’d be a lot better to have all of this information in a separate database and to use some sort of ref tag (possibly ref:Chido) on OSM buildings to link them to that database. That would allow you to keep razed/destroyed buildings in your own dataset (if you so desire) even after they’ve been removed from OSM. It would also allow you to use a different license for that database (if you want that).
This classification was build on this BAR methodology that uses: none - minimal - significant - complete
I think indeed that the none it is as you said; “a tag to indicate that the building/area was assessed and no damage was seen (as opposed to not knowing if a building was not yet assessed).”
I think I don’t agree in doing away the distinction between minimal and none because with minimal there is actually damage.
For complete, what about using destroyed:building=* ? I the wiki describes demolished:building=* rather as intentional removal which damage not necessarily is. In a lot of cases people will reconstruct it over time.
Hi @Friendly_Ghost,
Thanks for sharing your thoughts - much appreciated!
I think I wrote above my opinion on why I think this data has its place in OpenStreetMap, but I guess if most of the people here disagree with me, this should as well be clearly documented on the wiki. So I’m curious to hear from more of you on this point…
Hi @Friendly_Ghost ,
I was just rereading your message and I think I only now fully understood what you meant. Apologies.
I think it should indeed not be the goal to keep the non-existing buildings on OSM. So yes I think you are absolutely right in that. I agree that if a building is not existing anymore, it should not stay on OSM - and it could indeed further exist in another database.
However during eg. an earthquake a building does not just disapear, it is rather damaged and will be rebuild in a lot of situations - in which case it probably should stay on OSM - or in other situations the leftovers will be removed and it should than at that point be removed from OSM as well.
I am late jumping into this thread I had missed and just a quick reply as I am currently on a short trip with only my phone.
I tried to explain the use of (part of) the BAR methodology and the tagging I adopted in both the blog post and the OSM wiki page mentioned by @Jorieke_V . It does not interfere with the usual OSM tags, e.g. building. I am not convinced by the use of damage:building or destroyed:building for disasters, because they are not event-specific and need to be updated, which is not the case with the ones I adopted: they can remain and another future disaster can also be documented this way for the same buildings (of course if they are only damaged and not destroyed).
As explained by Jorieke, there is no existing database alternative to store this kind of data, and actually OSM is pretty good for the purpose IMO. Flexible and collaborative, but even regarding the long-term perspective someone mentioned. Keep in mind that OSM keeps track of the entire history of edits. It is not yet super easy to manage (even I think this might change soon with the new OSM format files and tools like those provided by HeiGIT), but with a bit of mastery of OSM history files you can process this in PostgreSQL.
Putting a tag for none and minimal is important IMO to avoid any doubt regarding the structure is undamaged or just not evaluated (this is an issue with the Copernicus EMSR data I mention, IMO). Also, minimal is the name of the class, but if you look at the BAR definition and the example provided, the structure has lost part of its roof, which can even make it unusable.
Regarding the completely damaged or destroyed buildings, yes, of course they will likely be replaced in the future, but as of now, in the most recent imagery available for OSM, they have not disappeared, but stand as piles of rubble, and they may remain for quite some time. In Haiti, destroyed buildings stood for years after the 2010 earthquake. We could use destroyed:building in addition of the damage tag, easy to add though.
For example, mapping whether given building was damaged during WW II or WW I or cyclone 40 years ago or earthquake 5 years ago or flood 150 years ago seems clearly out of scope for OSM for me.
Thanks everyone already for all your inputs. What do you think of the following proposal for mapping damaged buildings on OSM:
building=* damage=minimal OR damage=significant OR damage=complete source:damage=* (which will be most often the imagery used or survey) source:damage:date=YYYY-MM-DD
If the damage is because of a specific event the tag damage:nameofevent:event=GLIDE numbercould be added.
When a building is than removed fully - when there is even no rubble anymore to see on the ground - I would propose for these damage tags to be removed and the building can than be retagged with the destroyed:building=* or just deleted from OSM. Of course if the damage is repaired and the building being reconstructed, the damage tags should also disappear from OSM.
little to none, this is nearly the same thing (was: may be more fitting for objects that were moved/temporarily dismantled etc, which may happen even with buildings)
ruins: is more fitting if there are some still present remains, was: and destroyed: includes cases where no remains whatsoever remain
I have just come across this forum thread that might be relevant. I’m linking it here for reference at the moment, I have not fully read through it to completely understand it.